{"id":19734800,"url":"https://github.com/bhheo/ab_distillation","last_synced_at":"2025-04-30T04:31:14.256Z","repository":{"id":85123942,"uuid":"158494941","full_name":"bhheo/AB_distillation","owner":"bhheo","description":"Knowledge Transfer via Distillation of Activation Boundaries Formed by Hidden Neurons (AAAI 2019)","archived":false,"fork":false,"pushed_at":"2019-09-09T11:04:03.000Z","size":15774,"stargazers_count":103,"open_issues_count":0,"forks_count":18,"subscribers_count":5,"default_branch":"master","last_synced_at":"2024-05-22T19:35:56.123Z","etag":null,"topics":["knowledge-distillation","knowledge-transfer","network-compression","teacher-student-learning","transfer-learning"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/bhheo.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null}},"created_at":"2018-11-21T05:21:58.000Z","updated_at":"2024-01-04T16:28:15.000Z","dependencies_parsed_at":"2023-03-13T04:22:43.196Z","dependency_job_id":null,"html_url":"https://github.com/bhheo/AB_distillation","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bhheo%2FAB_distillation","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bhheo%2FAB_distillation/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bhheo%2FAB_distillation/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bhheo%2FAB_distillation/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/bhheo","download_url":"https://codeload.github.com/bhheo/AB_distillation/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":224198014,"owners_count":17271999,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["knowledge-distillation","knowledge-transfer","network-compression","teacher-student-learning","transfer-learning"],"created_at":"2024-11-12T00:39:36.558Z","updated_at":"2024-11-12T00:39:37.300Z","avatar_url":"https://github.com/bhheo.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Knowledge Transfer via Distillation of Activation Boundaries Formed by Hidden Neurons\n\nOfficial Pytorch implementation of paper:\n\n[Knowledge Transfer via Distillation of Activation Boundaries Formed by Hidden Neurons](https://arxiv.org/abs/1811.03233) (AAAI 2019).\n\nSlides and poster are available on [homepage](https://sites.google.com/view/byeongho-heo/home)\n\n## Environment\nPython 3.6, Pytorch 0.4.1, Torchvision\n\n\n## Knowledge distillation [(CIFAR-10)](https://www.cs.toronto.edu/~kriz/cifar.html) \n\ncifar10_AB_distillation.py\n\n\\\nDistillation from WRN 22-4 (teacher) to WRN 16-2 (student) on CIFAR-10 dataset.\n\nPre-trained teacher network (WRN 22-4) is included. Just run the code.\n\n## Transfer learning [(MIT_scenes)](http://web.mit.edu/torralba/www/indoor.html) \n\nMITscenes_AB_distillation.py \n\n\\\nTransfer learning from ImageNet pre-trained model (teacher) to randomly initialized model (student).\n\nTeacher : ImageNet pre-trained ResNet 50\n\nStudent : MobileNet or MobileNetV2 (randomly initialized model)\n\nPlease change base learning rate to 0.1 for MobileNetV2.\n\n\\\nMIT_scenes dataset should be arranged for Torchvision ImageFolder function.\n\n\nTrain set :\n`$dataset_path / train / $class_name / $image_name `\n\nTest set :\n`$dataset_path / test / $class_name / $image name`\n\n\nand run with dataset path.\n\nMobileNet\n```\npython MITscenes_AB_distillation.py --data_root $dataset_path\n```\n\nMobileNet V2\n```\npython MITscenes_AB_distillation.py --data_root $dataset_path --network mobilenetV2\n```\n## Other implementations\nTensorflow: https://github.com/sseung0703/Knowledge_distillation_methods_wtih_Tensorflow\n\n## Citation\n\n```\n@inproceedings{ABdistill,\n\ttitle = {Knowledge Transfer via Distillation of Activation Boundaries Formed by Hidden Neurons},\n\tauthor = {Byeongho Heo, Minsik Lee, Sangdoo Yun, Jin Young Choi},\n\tbooktitle = {AAAI Conference on Artificial Intelligence (AAAI)},\n\tyear = {2019}\n}\n```\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbhheo%2Fab_distillation","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fbhheo%2Fab_distillation","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbhheo%2Fab_distillation/lists"}